---
title: ✍️ Don’t throw the slop grenade
---

Saving 20 minutes could cost your team 40. Five things to agree on as a team. ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­    ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­  

| View in browser Issue #15 October 8, 2026 Say AI saves someone 20 minutes on a review. They send it to eight people. Each of those people spends five minutes working out what they’re being asked to do. That’s 40 minutes spent, so one person could save 20.   Frank Wolf, Chief Strategy Officer at Staffbase, knows that cost firsthand. He sent AI feedback to his own team with good intentions, and it didn’t land the way he expected. In this issue, he talks through what happened and shares five ways to avoid doing the same to your team.   Off to Frank! — Kim Pop quiz ↘ What share of employees rely on AI output without verifying it? 17% 37% 57% Scroll to the bottom for the answer. Frank Wolf, Chief Strategy Officer and Co-founder, Staffbase Here’sI threw a slop grenade at my own team, with good intentions. Here’s what I learned, and five ideas for your team.   A few weeks ago, my team launched a new page on our website. I reviewed it together with AI and gave it a lot of context: who the audience is, what the page should do, what we had discussed before. And honestly, the feedback was really good. It had a clear point of view, and there were ideas in it I would not have had myself.   It was also a lot. Long, very detailed, and when I read it again, not everything in it was correct. But the good ideas were good, so I put pretty much all of it into a message to my team. I told them clearly that this was AI feedback.   I felt they were really upset. For me this was a bit of a shock, because I had only good intentions. Looking back, I think I can see what happened.   While I gave the AI a lot of context, my team thought I gave almost none of my own judgment. I had thrown a slop grenade at my own team.   Lazy work used to look like nothing Before AI, lazy work mostly looked like nothing. Somebody didn’t know a topic or didn’t want to do the work, and so the work simply didn’t happen. That was bad, but at least nobody else had to clean it up. Bloated decks existed before, of course, but making them took effort, and effort was kind of a natural filter.   That filter is gone. Today lazy work lands as a 30-page report with a friendly “please review.” Shopify CEO Tobi Lütke said recently on Shane Parrish's podcast The Knowledge Project that the failure case of lazy work today is too much output. Inside Shopify they call these slop grenades, a term he credits to Harry Brundage.   I see it every week in slides. Simple ideas now come as 10- or 15-page decks, often from people who never had to make slides before and, for the first time, can make something beautiful. I really get it, it feels like a superpower. But now we can make a polished deck before deciding whether the idea needs one.   And it is not just me: in new, still preliminary research from Stanford’s Social Media Lab and BetterUp, more than half of the US desk workers surveyed said they had sent AI work they themselves consider low-effort or unhelpful.   The higher you sit, the bigger the blast When a colleague sends you 20 AI suggestions, you can ignore half of them. But when a co-founder sends them, every point becomes a to-do. So when my feedback creates real work, I also have to say what moves, what we drop or delay to make room. Otherwise my curiosity just becomes my team’s to-do list.   Do the math once. Say AI saves me 20 minutes on a review, and I send it to eight people. If each of them spends five minutes just figuring out what I actually want, the team spent 40 minutes so I could save 20.   For internal comms teams, the handoff happens twice: from the leader to your desk, and from your desk to the whole workforce. Internal comms was always a gatekeeper, making sure things are well understood, and with AI this role becomes really critical.   A leader’s AI draft can be long and seemingly well written, in parts maybe better than you would write it yourself, and still the point and the next steps are hard to find. But employees can’t send a company-wide email back, so comms has to push back for them. ✍️ Defusing the grenade   There are different ways to deal with this. As always, I think the first and most obvious one is to talk about it, in your team or on a company level.   Having a name helps here. “Slop grenade” might feel too harsh for some, but it works like a company value that really sticks: It gives a behavior a name everybody remembers. In this case, it is a behavior you want to see less of, and that is a great start.   Beyond that, here are five more concrete ideas for everyday work: Decide what “good enough” means before you ask AI for feedback. If you ask how to improve something, AI will always find improvements, even when none are needed. Ask instead whether the work meets the goals you set, and be ready to accept “yes, ship it.” Lead with your own judgment, not the AI’s output. Open with something like: “Based on what I found, here’s my recommendation and what I need from you. Do you agree?” Then treat the AI material as backup for anyone who wants to dig deeper. Consider sharing the prompt instead of the result. With speech-to-text, my prompts are long and full of context, doubts and priorities. That is my real thinking, and my team often needs it more than the polished answer. Talk about limits. Three slides is my maximum to make a case, and very often one is enough. It doesn’t even have to be a slide. Make it safe for people to push back. If someone asks, “Which parts do you want us to act on?” your request wasn’t clear enough yet. Answer with a sharper ask instead of defending the document. So what would I send today? Not the whole review. I would pick the few points I really believe in and write them in my own words. Leaving things out was the part of the work I had skipped.   Claude can build me the grenade. Throwing it is still my decision.   \* \* \*   If this resonates, or you’re seeing this differently, I’d like to hear it. Reply to this email. Or reach out on LinkedIn. 📚 Reading list The Knowledge Project Podcast   Tobi Lütke: AI agents, better decisions, and the future of work The conversation behind this issue. Shopify CEO Tobi Lütke speaks with host Shane Parrish about AI agents, decision-making, and the future of work. He covers how Shopify works with an AI colleague called River, and why lazy work today means too much output. Full episode → Association for Computing Machinery   From workslop to work: how AI explanations reconfigure human judgment in creative tasks Spotting workslop is easy. Admitting to it is harder. In this preprint, more than half of the US desk workers surveyed said some of the AI work they send may be low-effort. The researchers also examine what leads people to create it and what it costs the teams on the receiving end. Full study → Harvard Business Review AI doesn’t reduce work — it intensifies it Berkeley researchers followed one tech company for eight months and found that it intensifies work instead. They argue that leaving AI habits to individuals isn’t enough, and that teams need shared norms. Read more → The answer is c. 57%. The same research found 48% of employees use AI in ways that go against company policy. The fix isn’t less AI; it’s clearer expectations about what people can use AI for and what they should always check. Here’s how to close the gap. Did someone forward you this newsletter? Subscribe to The Workplace Rewrite to receive the next issue! Staffbase Inc., 408 Broadway St, New York, NY 10013, United States Unsubscribe Manage preferences Site Notice Privacy Policy |
| --- |